oracle表连接------>排序合并连接(Merge Sort Join)

转载：http://blog.csdn.net/waterxcfg304/article/details/25873265

排序合并连接 (Sort Merge Join)是一种两个表在做连接时用排序操作(Sort)和合并操作(Merge)来得到连接结果集的连接方法。

对于排序合并连接的优缺点及适用场景如下：

a,通常情况下，排序合并连接的执行效率远不如哈希连接，但前者的使用范围更广，因为哈希连接只能用于等值连接条件，而排序合并连接还能用于其他连接条件(如<,<=,>.>=)

b,通常情况下，排序合并连接并不适合OLTP类型的系统，其本质原因是对于因为OLTP类型系统而言，排序是非常昂贵的操作，当然，如果能避免排序操作就例外了。

oracle表之间的连接之排序合并连接(Merge Sort Join)，其特点如下：

1，驱动表和被驱动表都是最多只被访问一次。

2，排序合并连接的表无驱动顺序。

3，排序合并连接的表需要排序，用到SORT_AREA_SIZE。

4，排序合并连接不适用于的连接条件是：不等于<>，like,其中大于>，小于<，大于等于>=，小于等于<=，是可以适用于排序合并连接

5，排序合并连接，如果有索引就可以排除排序。

下面我来做个实验来证实如上的结论：

具体的测试基础表请查看本人Blog 如下链接：

oracle表连接之----〉嵌套循环(Nested Loops Join)

1，驱动表和被驱动表的访问次数：

sql> select /*+ ordered use_merge(t2)*/ * from t1,t2 where t1.id=t2.t1_id;

sql> select sql_id,child_number,sql_text from v$sql where sql_text like '%use_merge%';

sql_ID CHILD_NUMBER sql_TEXT
------------- ------------ --------------------------------------------------------------------------------
85u4h9hfqa5ar 0 select sql_id,sql_text from v$sql where sql_text like '%use_merg
6xph9fhapys39 0 select /*+ ordered use_merge(t2)*/ * from t1,t2 where t1.id=t2.t1_id

sql> select * from table(dbms_xplan.display_cursor('6xph9fhapys39','allstats last'));

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
sql_ID 6xph9fhapys39,child number 0
-------------------------------------
select /*+ ordered use_merge(t2)*/ * from t1,t2 where t1.id=t2.t1_id
Plan hash value: 412793182
--------------------------------------------------------------------------------
| Id | Operation | Name |Starts| E-Rows | A-Rows | A-Time | Buf
--------------------------------------------------------------------------------
| 1 | MERGE JOIN | | 1 | 100 | 100 |00:00:00.07 |
| 2 | SORT JOIN | | 1 | 100 | 100 |00:00:00.01 |
| 3 | TABLE ACCESS FULL| T1 |1| 100 | 100 |00:00:00.01 |
|* 4 | SORT JOIN | | 100 | 100K| 100 |00:00:00.07 |
| 5 | TABLE ACCESS FULL| T2 | 1| 100K| 100K|00:00:00.01 |
--------------------------------------------------------------------------------
Predicate information (identified by operation id):
---------------------------------------------------
4 - access("T1"."ID"="T2"."T1_ID")

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
filter("T1"."ID"="T2"."T1_ID")
Note
-----
- dynamic sampling used for this statement

26 rows selected

从上面的实验可以看出排序合并连接和HASH连接时一样的，T1和T2 表都只会被访问0次或者1次。

select /*+ ordered use_merge(t2)*/ * from t1,t2 where t1.id=t2.t1_id and 1=2;此语句T1和T2表就会是被访问0次。自己可以做试验测试下。

总结：排序合并连接根本就没有驱动和被驱动表的概念，而嵌套循环连接和哈希连接就要考虑驱动和被驱动表的情况！！

2，排序合并的表的驱动顺序

下面是T1为驱动表的执行计划

select /*+ leading(t1) use_merge(t2)*/ * from t1,t2 where t1.id=t2.t1_id and t1.num=20;
select sql_id,sql_text from v$sql where sql_text like '%from t1,t2 where t1.id=t2.t1_id and t1.num=20%';
sql> select * from table(dbms_xplan.display_cursor('8z4jvhnnfhxyf','allstats last'));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
sql_ID 8z4jvhnnfhxyf,child number 0
-------------------------------------
select /*+ leading(t1) use_merge(t2)*/ * from t1,t2 where t1.id=t2.t1_id and t1.num=20

Plan hash value: 412793182

-----------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | OMem | 1Mem | Used-Mem |
-----------------------------------------------------------------------------------------------------------------
| 1 | MERGE JOIN | | 1 | 1 | 1 |00:00:00.58 | 3462 | | | |
| 2 | SORT JOIN | | 1 | 1 | 1 |00:00:00.01 | 6 | 2048 | 2048 |2048(0)|

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
|* 3 | TABLE ACCESS FULL| T1 | 1 | 1 | 1 |00:00:00.01 | 6 | | | |
|* 4 | SORT JOIN | | 1 | 100K| 1 |00:00:00.58 | 3456 | 14M| 1490K|12M(0)|
| 5 | TABLE ACCESS FULL| T2 | 1 | 100K| 100K|00:00:00.01 | 3456 | | | |
-----------------------------------------------------------------------------------------------------------------

Predicate information (identified by operation id):
---------------------------------------------------

3 - filter("T1"."NUM"=20)
4 - access("T1"."ID"="T2"."T1_ID")
filter("T1"."ID"="T2"."T1_ID")

23 rows selected.

Elapsed: 00:00:00.01

下面是T2为驱动表的执行计划：

sql> select * from table(dbms_xplan.display_cursor('bxydvw58bhczf',43); font-family:Arial; font-size:14px; line-height:26px"> PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
sql_ID bxydvw58bhczf,child number 0
-------------------------------------
select /*+ leading(t2) use_merge(t1)*/ * from t1,43); font-family:Arial; font-size:14px; line-height:26px"> Plan hash value: 1792967693

-----------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | OMem | 1Mem | Used-Mem |
-----------------------------------------------------------------------------------------------------------------
| 1 | MERGE JOIN | | 1 | 1 | 1 |00:00:02.20 | 3462 | | | |
| 2 | SORT JOIN | | 1 | 100K| 21 |00:00:02.20 | 3456 | 14M| 1490K|12M(0)|

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| 3 | TABLE ACCESS FULL| T2 | 1 | 100K| 100K|00:00:00.10 | 3456 | | | |
|* 4 | SORT JOIN | | 21 | 1 | 1 |00:00:00.01 | 6 | 2048 | 2048 |2048(0)|
|* 5 | TABLE ACCESS FULL| T1 | 1 | 1 | 1 |00:00:00.01 | 6 | | | |
-----------------------------------------------------------------------------------------------------------------

4 - access("T1"."ID"="T2"."T1_ID")
filter("T1"."ID"="T2"."T1_ID")
5 - filter("T1"."NUM"=20)

Elapsed: 00:00:00.85

从上面的两个执行计划可以看出，无论T1表示驱动表还是被驱动表，效果都是一样的，排序的尺寸一个是2048+12M，一个是12M+2048。

结论：排序合并连接没有驱动的概念，无论哪个表再前面都无所谓。

3，排序合并连接的限制

sql〉explain plan for select /*+ leading(t1) use_merge(t2)*/ * from t1,t2 where t1.id<>t2.t1_id and t1.num=20;

sql> select * from table(dbms_xplan.display);

---------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%cpu)| Time |
---------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 5000 | 1083K| 82709 (1)| 00:15:10 |
| 1 | nesTED LOOPS | | 5000 | 1083K| 82709 (1)| 00:15:10 |
| 2 | TABLE ACCESS FULL| T2 | 100K| 10M| 710 (1)| 00:00:08 |
|* 3 | TABLE ACCESS FULL| T1 | 1 | 107 | 1 (0)| 00:00:01 |
---------------------------------------------------------------------------

3 - filter("T1"."NUM"=20 AND TO_CHAR("T1"."ID") LIKE
TO_CHAR("T2"."T1_ID"))

16 rows selected.

从上面的执行计划可以看出，优化器走的是nesTED LOOPS JOIN。

sql> explain plan for select /*+ leading(t1) use_merge(t2)*/ * from t1,t2 where t1.id>t2.t1_id and t1.num=20;

Explained.

Elapsed: 00:00:00.01
sql> select * from table(dbms_xplan.display);

------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%cpu)| Time |
------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 5000 | 1083K| | 5080 (1)| 00:00:56 |
| 1 | MERGE JOIN | | 5000 | 1083K| | 5080 (1)| 00:00:56 |
| 2 | SORT JOIN | | 1 | 107 | | 4 (25)| 00:00:01 |
|* 3 | TABLE ACCESS FULL| T1 | 1 | 107 | | 3 (0)| 00:00:01 |
|* 4 | SORT JOIN | | 100K| 10M| 25M| 5076 (1)| 00:00:56 |
| 5 | TABLE ACCESS FULL| T2 | 100K| 10M| | 710 (1)| 00:00:08 |

3 - filter("T1"."NUM"=20)
4 - access(INTERNAL_FUNCTION("T1"."ID")>INTERNAL_FUNCTION("T2"."T1_ID"))
filter(INTERNAL_FUNCTION("T1"."ID")>INTERNAL_FUNCTION("T2"."T1_ID"))

19 rows selected.

同理可以实验得出：排序合并连接不适用于的连接条件是：不等于<>，like,51)">其中大于>，小于<，大于等于>=，小于等于<=，是可以适用于排序合并连接

oracle表连接------>排序合并连接(Merge Sort Join)

相关文章